A revolutionary novelty in cervical cancer screening at the University of Debrecen

University

Thanks to the joint development of the University of Debrecen, Delta Services and the Sightspot Network, it is possible to diagnose cervical cancer faster and more reliably, so that the treatment of the disease can be started in time. An imaging system based on in-depth learning technology identifies cancer cells.

Cervical cancer screening could be revolutionized by the joint development of the consortium, the Faculty of Informatics at the University of Debrecen said at a briefing on the diagnostic system on Friday. The research of artificial intelligence is central to the scientific work of the faculty, and the development of a diagnostic system fits in perfectly with this. The results of the screening tests currently have to wait 2-4 weeks. However, thanks to the Hungarian development, the new software used for filtering the samples, supported by artificial intelligence, helps to evaluate the results faster.

András Hajdu, Dean of the Faculty of Informatics of the University of Debrecen, emphasized that the prototype was completed and could be used in practice.

The aim is to introduce the new diagnostic system into clinical practice. However, this requires another international test and evaluation. With artificial intelligence, results can be achieved cost-effectively and quickly. And thanks to early detection, doctors can perform significantly more life-saving interventions. In the project, great emphasis was placed on manual work, supervised machine teaching. The database will be used not only in cytology but also in other fields,

– the faculty leader added.

László Vidra, Managing Director of Delta Systems Kft., Emphasized that the method can significantly relieve the burden on healthcare, as it is able to detect cells showing changes in digitized smears without the use of human labor.

Digitization is key to competitiveness and the project can significantly improve skills shortages. We can play a catalytic role in a particularly important area, as the development of digitalisation is also a primary goal for the national economy. In connection with this, we used deep learning technology with the help of artificial intelligence. The cooperation between the university and the profession, the system integrator that helps to enter the market, is exemplary, which also indicates that a forward-looking ecosystem is emerging,

– the executive emphasized.

Ilona Kovács, head of the Department of Pathology at the Clinical Center of the University of Debrecen, a cytopathologist, noted that software had been developed that was able to provide quality control for the selection of abnormal cells.

A medical team worked on the digitization process, image evaluation, and sample selection during the training phase of the system. Based on data from the IT team, it processed a total of more than 264,000 cell images out of ten thousand recorded smears for machine learning and testing. If there is a change in strategy in cervical cancer screening, the development could live up to its promise,

– the chief medical officer emphasized.

Balázs Harangi, an associate professor at the Faculty of Informatics of the University of Debrecen, said that according to previous experience, the system is currently operating with ninety-three percent accuracy at the cellular level.

The work of cytologists in collecting smears and digitizing was essential. Our main task was to extract the cells from the collected smears using traditional and advanced machine image processing techniques. One hundred thousand times two hundred thousand pixel resolution images were taken in seven gigabytes each. We want to run a test that proves that the system works and can be used confidently in practice,

– he said.

The development was implemented through a call for proposals entitled “Competitiveness and Excellence in R&D” published within the framework of the Széchenyi 2020 program.

 

hirek.unideb.hu

Leave a Reply

Your email address will not be published. Required fields are marked *